The placement of radioactive isotopes directly into a cancer of the uterine cervix or vagina in order to eradicate the cancer, requires insertion of a hollow applicator into the tumor. Though previously plain x-ray film depicted the position of the applicator, and was used to calculate the dose to deliver to an area around the applicator, 3D CT revolutionized tumor visualization in the 1990s. However, MRI more clearly defines gynecologic tumors, though the application of MRI to gynecologic brachytherapy is in its infancy. Research in the last five years has shown that using CT alone overestimates the extent of normal tissue that surrounds the cancer compared to MR. In comparison MR allows the radiation oncologist to draw tighter boundaries around the target volume that should receive the highest radiation dose, while sparing larger amounts of surrounding normal tissue. Current practice at our hospital after applicator insertion in the operating room is to obtain a 3D image to guide radiation treatment planning. Patients with large tumors undergo both a CT and an MRI. Commercial software aligns the MRI and CT to each other, and is used to outline target cancer and surrounding organs at risk. A treatment plan is generated and radiation is delivered to the patient using this plan. Current alignment and contouring methods require significant expert input from the radiation oncologist and physicist while the patient waits for treatment under anesthesia. This project will improve the efficiency of this workflow using advanced image analysis methods as follows. 1) We will develop an EM-ESP segmenter to simultaneously segment registered MR and CT scans for gynecologic brachytherapy planning. This segmenter will construct and use joint prior probability distributions on image intensities and the geometry of key anatomic structures relative to the brachytherapy applicator, which is clearly visible in the images. 2) We will apply the EM-ESP segmenter retrospectively on a database of 25 patients who were treated during 2004-2006 using MR guided brachytherapy by Dr. Akila Viswanathan, Chief of Gynecologic Radiation Oncology at our hospital. Contours were drawn manually on these scans, and we will test our method against these. The result of this project will be an algorithm that can automatically delineate the cervix, the bladder, the rectum, the sigmoid colon, and the small bowel in MR and CT images for gynecologic brachytherapy. To the best of our knowledge, the proposed algorithm will be the first to provide this functionality. ( ( ! PUBLIC HEALTH RELEVANCE: Magnetic resonance images have proven to identify the extent of cancer in gynecologic cancer better than other imaging modalities. The better the cancer can be differentiated from the surrounding bladder, rectum, and small bowel, the more accurately high doses of radiation can be given to kill the tumor cells while minimizing the toxic effects of radiation on normal tissue. This project will develop algorithms will automatically identify the boundaries of the cancer and surrounding normal organs of the human female pelvis so that MR-guided radiation therapy can be carried out efficiently and accurately.